Methods
Literature Search
In order to identify any existing instruments/guidelines which can be used for assessing the suitability of sidewalks and roads for walking and bicycling, we searched professional literature that encompassed public health, transportation engineering, and transportation planning technical reports, peer-reviewed literature, and internet websites. We identified seven instruments/guidelines related to the assessment of the suitability of pedestrian environments, and 22 instruments/guidelines related to the assessment of the suitability of bicycling environments. Almost all of them were developed by transportation planners and traffic engineers attempting to incorporate pedestrian or bicycling needs within the general transportation planning process. We found no “gold standard” assessment instruments in the literature, therefore, we statistically compared only the quantitative instruments to identify reliable, valid, and relatively simple instruments for collecting objective and easily available data on walking and bicycling suitability.
None of the pedestrian suitability assessment instruments we identified met our primary criteria of collecting objective, quantifiable, and easily available data for sidewalk assessments. Table 1 lists the seven instruments/guidelines we found and the reasons we excluded them.
Table 1. Pedestrian Suitability Assessment Instruments Identified from Literature Search
| # | Author | Method Name | Location Piloted | Reasons Excluded |
| 1 | Partnership for a Walkable America (1997) | Walkable America Checklist | Nationally promoted | No objective, quantifiable measures that can be used for systematic data collection. |
| 2 | D. Bonk, S. Luxenberg (1997) | Pedestrian Facility Survey | Chapel Hill and Carrboro, NC | No objective, quantifiable measures that can be used for systematic data collection. |
| 3 | L. Dixon (1996) | Pedestrian Level of Service | Gainesville, FL | Requires motor vehicle LOS. Requires complicated intersection design considerations. |
| 4 | S. Sarkar (1995) | Pedestrian Service Levels and Quality of Service | Unknown | Requires long-term familiarity with segments to provide “average” values for characteristics |
| 5 | D. Mozer (1994) | LOS Pedestrians | Unknown | Requires volume of pedestrian traffic |
| 6 | C. Bradshaw (1993) | Neighborhood Walkability | Ottawa, CA | Measures are not feasibly operationalized |
| 7 | Y. Tanaboriboon and J. Guyano (1989) | Pedestrian Level of Service | Bangkok, Thailand | Measures flow rate of pedestrians, but not environmental characteristics |
Note: Pedestrian Level of Service” (TRB Paper No. 01-0511) by Bruce Landis was published in 2001 and not included in this study.
To develop a new walking suitability assessment instrument, we identified variables from the literature that characterized the pedestrian environment, including: type of sidewalks present; sidewalk surface material; sidewalk surface condition; sidewalk width; buffer width; curb ramp presence; and street lighting. These variables were prioritized and weighted by importance for safe and accessible walking based on published pedestrian instruments.
To calculate the suitability of the sidewalk, we developed an algorithm that summed the values for the sidewalk characteristics. Staff at the UNC-Chapel Hill Highway Safety Research Center reviewed the instrument and suggested additional improvements. To see if the scores reflected the quality of the road segments relative to one another, we pre-tested the instrument on sidewalks exhibiting a variety of walking conditions. The final instrument was prepared for use in Phase Two.
Bicycling Assessment Instrument
Out of the 22 assessment methods we identified, only 13 were quantitative instruments that assessed bicycling suitability of roads. One of the 13 instruments (Dixon, 1995) required motor vehicle level of service data which were not available to us for every road segment. Therefore, we excluded that instrument from analysis and compared 12 of the bicycle suitability assessment instruments. Table 2 provides a list of the exclusion criteria for the ten additional instruments/guidelines not included in the study.
Table 2. Ten Bicycle Assessment Instruments Not Included in Analysis
| # | Author | Instrument Name | Location Piloted | Reason Not Included |
| 1 | Institution of Highways and Transportation (1998) | Guidelines for Cycle Audit and Cycle Review | London, England | Pilot tool promoted via internet |
| 2 | William Barber (1997) | Measuring Bicycle Accessibility | OR | No suitability assessment formula developed |
| 3 | Main Roads (1997) | Safety Audit Checklist for Dual-Use Paths | Western Australia | No quantitative data collection |
| 4 | Linda Dixon (1995) | Gainesville Bike Level of Service and Performance Measures (LOS) | FL | Requires motor vehicle LOS(1) |
| 5 | Slade McCalip (1995) | Rural Suitability Rating Index (RSRI) | NC | Only for rural roads |
| 6 | Hein Botma (1995) | Level of Service for Bicycle Paths and Pedestrian-Bicycle Paths | Netherlands | Assesses only off-road bicycle paths |
| 7 | David Mozer (1994) | LOS Bicycles | Unknown | Requires volume of bicycle traffic(2) |
| 8 | Alex Sorton (1994) | Determining Road Compatibility for Bicyclists | IL | Revised by Sorton and Walsh (see BSLRC instrument) |
| 9 | Hakkert and Pistiner (1988) | Environmental Quality and Safety Assessment of Residential Streets | Haifa, Israel | Not a suitability assessment model |
| 10 | Jeff Davis (1987) | Bicycle Safety Index Rating (BSIR) | TN | Revised by Epperson-Davis(see RCI instrument) |
(1) Motor Vehicle Level of Service (FHWA Highway Capacity Manual designation) was not obtainable for each road segment within the study timeframe and necessitated the exclusion of the Dixon instrument from analysis.
(2) Impractical measure for areas without dense bicycle-riding population.
In order to statistically compare the 12 assessment instruments, we identified Greenville, NC as the location for data collection because it was the site for one of the regional North Carolina Cardiovascular Health (CVH) Program offices. In addition, the Greenville Metropolitan Planning Organization (MPO) transportation planner had compiled a dataset of road characteristics for 296 urban and rural road segments that ranged from 0.1 to 2.0 miles in length. Roads were segmented at major intersections or where objective road characteristics changed, such as number of travel lanes or posted speed limit. We randomly sampled 30 road segments from the MPO dataset.
The number of variables required to complete each of the 12 assessment instruments ranged from four to thirty. Many of the bicycle suitability assessment instruments included similar variables in their scoring algorithms We developed a primary data collection form to assess the 59 variables necessary to compute suitability across each of the twelve instruments. Although secondary data sources provided outside lane width, early verification suggested that the data were often inaccurate. Therefore, we used a steel measuring tape to measure the width of the outside travel lane. We used secondary data for the traffic volume, also know as annual average daily traffic (AADT).
Two people collected data, one with experience as a commuting cyclist and the other as a recreational cyclist. They traveled the 30 road segments together in the same motor vehicle from 8:00am until 7:00pm on a non-holiday weekday during fair weather in the Fall of 1999. From within the car, they independently collected the data on separate forms, only leaving the car to measure the width of the travel lanes. Because AADT was not available for one of the road segments, the sample was reduced to 29 road segments. Microsoft Excel 2000 was used to compute each instrument’s algorithm to produce a total of twelve suitability scores for each segment. Since one instrument’s calculations required us to compute the natural log of a zero value, we dropped that segment from the analysis for all twelve instruments, leaving the final analysis sample in Phase 1 as 28 road segments. We ensured the quality of our data collection for each instrument by testing the inter-coder reliability for the research staff using intra-class correlations.
In order to select one of the twelve instruments when no assessment instrument exists in the literature as a “gold standard” for criterion-related validity comparison, we correlated each one with all the remaining instruments (a proxy “standard”). The statistic we used was an item-total correlation applying a Cronbach’s alpha coefficient using SAS Version 8.0. We used the results of this analysis (manuscript submitted for publication) to identify one bicycling instrument for Phase Two. The N. Eddy method that used adjective categories for pavement condition was selected as most highly correlated and had the added benefit of being one of the most simple to use.